fix: Fix gemma models broken by HF update

[yfw]

What does this PR do ?

Gemma models were broken in NeMo-RL and vLLM by this HF commit: huggingface/transformers#37033. To address this:

• Patch vllm to include gemma change from this PR: [Bugfix] Update multimodel models mapping to fit new checkpoint after Transformers v4.52 vllm-project/vllm#19151
• Fix Gemma parallelism in NeMo-RL that was broken by model_prefix change

Issues

List issues that this PR closes (syntax):

Usage

• You can potentially add a usage example below

# Add a code snippet demonstrating how to use this

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you run the unit tests and functional tests locally? Visit our Testing Guide for how to run tests
• Did you add or update any necessary documentation? Visit our Document Development Guide for how to write, build and test the docs.

Additional Information

• ...